Setting up an etcd store for HADR in Db2
You must set up an etcd store to enable automated failover in a Db2 High Availability Disaster Recovery (HADR) configuration on Cloud Pak for Data.
Db2 HADR supports automated failover by using a mechanism called Governor. Governor relies on an etcd key-value store to keep track of HADR state information, such as the current leader (primary node). Automated failover is only supported between the primary and principal standby, so a single etcd endpoint must be shared between the two deployments.
Note: If using the built-in etcd store with
the etcd pod, create a network policy to allow
communication from all deployments with the etcd pod. See Creating a network policy for built-in etcd for details.
Important: For production deployments, a three-node etcd topology is required.
If etcd is installed on the same OpenShift® cluster, it should not be co-located on the same worker nodes as the Db2 deployments. This can be achieved using pod anti-affinity in the etcd deployment.
To set up an etcd store for HADR, choose from the following methods:
- You can use an external etcd store that can be a cloud-provided service, deployed on the OpenShift cluster, or hosted on-premises.
- Deploy etcd packaged with the Db2 installation
- Deploy etcd packaged by Bitnami for Db2
- Using the built-in Db2 etcd store (only non-production, single OpenShift cluster environments)